DeepReinforce, the AI lab founded by prominent NLP scholar Li Jiwei, has open-sourced Ornith-1.0, a family of four coding-agent models released under the MIT license. Built on top of Gemma 4 and Qwen 3.5, the suite spans both dense and mixture-of-experts (MoE) architectures. The 397B flagship model outperforms Claude Opus 4.7 on several programming benchmarks, including Terminal-Bench 2.1 (Claude Code).
